Back to EveryPatent.com
United States Patent |
6,240,462
|
Agraharam
,   et al.
|
May 29, 2001
|
System for providing enhanced grade of service for connections over a large
network
Abstract
We have realized that for a connection over a large data network, such as
an Internet connection that couples a web access device to a terminal
server connected to an applications server, delays associated with the
process of downloading large data files over such a connection, occur
primarily in the leg of the connection that couples the applications
server to a terminal server while little congestion is typically observed
in the leg of the connection that couples the Internet access device to
the terminal server. The aforementioned delay is reduced by establishing a
separate connection from the applications server to the Point Of Presence
server outside or independently of the backbone of the large data network
thereby allowing users to receive enhanced grade of service for file
transfer operations.
Inventors:
|
Agraharam; Sanjay (Marlboro, NJ);
DeSimone; Antonio (Ocean, NJ);
Kuthyar; Ashok K. (Holmdel, NJ);
Ramamurthy; Ram S. (Manalapan, NJ);
Sibal; Sandeep (Matawan, NJ)
|
Assignee:
|
AT&T (New York, NY)
|
Appl. No.:
|
950021 |
Filed:
|
October 14, 1997 |
Current U.S. Class: |
709/238; 370/238 |
Intern'l Class: |
G06F 013/00 |
Field of Search: |
395/200.48,200.49,200.47
709/218,219,217,239,238
379/221
370/238
|
References Cited
U.S. Patent Documents
5694549 | Dec., 1997 | Carlin et al. | 709/250.
|
5732078 | Mar., 1998 | Arango | 370/355.
|
5845073 | Dec., 1998 | Carlin et al. | 709/217.
|
5862339 | Jan., 1999 | Bonnaure et al. | 709/227.
|
5905872 | May., 1999 | DeSimone et al. | 709/245.
|
5933425 | Aug., 1999 | Iwata | 370/351.
|
5933490 | Aug., 1999 | White et al. | 379/221.
|
5940372 | Aug., 1999 | Bertin et al. | 370/238.
|
5953312 | Sep., 1999 | Crawley et al. | 370/218.
|
6084858 | Jul., 2000 | Matthews et al. | 370/238.
|
Other References
David Lewis et al.; Multi-Service Management in a Multi-Provider
Environment; Fifth IEE Conference on Telecommunications; pp. 292-296,
1995.
|
Primary Examiner: Coulter; Kenneth R.
Attorney, Agent or Firm: Kenyon & Kenyon
Claims
What is claimed is:
1. A method for transferring data between a web access device and an
applications server, wherein both, the web access device and the
applications server are coupled to a Point of Presence server in a
network, said method comprising the steps of:
a) routing commands from the web access device to the applications server
via the network;
b) downloading data from the applications server to the Point of Presence
server via a subnetwork having a predetermined quality of service, said
downloading step being performed after accessing a database to determine
the subnetwork via which the applications server is to download the data
to the web access device; and
c) selecting an optimum subnetwork from among a plurality of subnetworks,
said selection being made according to a set of predetermined criteria.
2. The method according to claim 1, wherein said downloading step includes
the step of:
mapping an IP address to a subnet address, wherein the subnet address is a
combination of the subnetwork and an address within the subnetwork for a
machine identified by the IP address.
3. The method according to claim 1, further comprising the step of:
accessing a server on the network on which is stored the database via which
any device on the network can find out the optimum subnetwork when
selecting a subnetwork to another device.
4. The method according to claim 1, wherein the predetermined criteria
comprises:
a) an ability of the available subnetworks to support the quality of
service desired;
b) a current traffic load on the available subnetworks;
c) a price/payment for each of the available subnetworks; and
d) time of day variations that the subnetworks experience.
5. The method according to claim 2, further comprising the step of storing
the optimum subnetwork's identity in a database that is accessible by a
server on the network.
6. The method according to claim 5, further comprising the step of
dynamically updating the database based on a plurality of predetermined
criteria.
7. The method according to claim 6, wherein the predetermined criteria
comprises:
a) an ability of the available subnetworks to support the quality of
service desired;
b) a current traffic load on the available subnetworks;
c) a price/payment for each of the available subnetworks; and
d) time of day variations that the subnetworks experience.
8. The method according to claim 2, further comprising the step of using a
default IP network in the event that no alternate subnetwork is available.
9. The method according to claim 1, further comprising the step of:
establishing a connection from the applications server to a circuit
switched network and from the circuit switched network to a dial-up side
of the Point Of Presence server.
10. The method according to claim 1, further comprising the step of:
establishing a connection from the applications server to a circuit
switched network and from the circuit switched network to a network side
of the Point Of Presence server.
11. The method according to claim 1, further comprising the step of:
establishing a connection from the applications server to a circuit
switched network and from the circuit switched network to a neighboring
server within the network, wherein the neighboring server includes a
server that is logically close to the Point Of Presence server.
12. The method according to claim 1, further comprising the step of:
establishing a connection from the applications server to a circuit
switched network to a server having a predetermined maximum number of
connections needed to reach the Point Of Presence server.
13. The method according to claim 1, wherein the alternate path is
established from a network selected from a group of networks comprised of
a) quality of service asynchronous transfer mode network, b) another IP
network that is RSVP enabled internally between routers, c) a Frame Relay
network, d) a POTS circuit switched network, and e) an ISDN network.
14. In a client-applications server networked environment, in which the
Internet Protocol is implemented and the client is connected to the
applications server through a terminal server and the client can request
data from the applications server by sending commands to the applications
server through the terminal server, a method for communicating between the
client and the server comprising the steps of:
a) separating a command path from a data path within the network; and
b) transmitting data to the client from the applications server along the
data path based upon a subnet address included in a command sent along the
command path wherein the data path is external to the client-applications
server network.
15. The method according to claim 14, wherein the command path comprises a
path taken by a request to download data from the client to the
applications server.
16. The method according to claim 14, wherein the data path is a
circuit-switched connection from the applications server to the terminal
server.
17. The method according to claim 14, wherein the data path is within the
plain old telephone switched infrastructure using PPP protocol.
18. The method according to claim 14, wherein the command path is within
the client-applications server network using IP protocol.
19. The method according to claim 14, further comprising the step of
embedding signaling information in an HTTP request by a proxy server.
20. A method for interacting with an applications server by a client
accessing the applications server via a Point Of Presence server,
comprising the steps of:
a) transmitting a client request using Internet protocol to the Point Of
Presence server;
b) routing the client request over a network to the applications server;
c) establishing an alternate path back to the Point Of Presence server
using a different protocol;
d) downloading information to the Point Of Presence server using the
alternate path and the different protocol; and
e) routing the information from the Point Of Presence server to the client
using the IP protocol.
21. The method according to claim 20, further comprising the step of:
establishing for said alternate path a connection from the applications
server to a circuit-switched network and from the circuit-switched network
to a dial-up side of the Point Of Presence server.
22. The method according to claim 20, further comprising the step of:
establishing for said alternate path a connection from the applications
server to a circuit-switched network and from the circuit-switched network
to a network side of the Point Of Presence server.
23. The method according to claim 20, further comprising the step of:
establishing for said alternate path a connection from the applications
server to a circuit-switched network and from the circuit-switched network
to a neighboring server within the network, wherein the neighboring server
includes a server that is logically close to the Point Of Presence server.
24. The method according to claim 20, further comprising the step of:
establishing a connection from the applications server via the
circuit-switched network to yet another server having a predetermined
maximum number of connections needed to reach the Point Of Presence
server.
25. A communications system for communicating data between an applications
server and a client device, the applications server and the client device
both being coupled to a first network which supports the IP protocol and
which has a plurality of routers, said system comprising:
a) a Point Of Presence server which includes (1) a plurality of dial-up
access ports, wherein one of the plurality of access ports is coupled to
the client device via a dial-up connection; and (2) a plurality of router
ports in said routers to which said Point Of Presence server to said
applications server are coupled; and
b) a database server coupled to the Point Of Presence server via said
router ports, said database server storing a list of alternate paths for
downloading data from the applications server to the Point Of Presence
server, said list of alternate paths being dynamically updated based on a
plurality of predetermined criteria.
26. The apparatus according to claim 25, wherein the alternate paths
include at least one path which is established on a second network and
which is selected from a group of networks which include a quality of
service asynchronous transfer mode network, a separate IP network that is
RSVP-enabled internally between routers, a frame relay network, a
circuit-switched POTS network and a circuit-switched ISDN network.
27. A method of communicating information between a network access device
and an applications server, comprising the steps of:
establishing a connection for communicating commands from said network
access device to said applications server, said connection being
established between said network access device and said applications
server via a terminal server which is coupled to said applications server
through a first communications network;
transmitting data which is destined for said network access device, from
said applications server to said terminal server via a second
communications network; and
accessing a database to identify said second communications network for
said data transmission, wherein said second communications network is
selected from a group of networks which include an ATM network, a frame
relay network and a circuit-switched network.
28. The method of claim 27 wherein said first communications network is a
packet-switched network and said second communications network is a
circuit-switched network.
Description
BACKGROUND OF THE INVENTION
The present invention relates generally to a method for providing enhanced
quality of service for networks, and more particularly to a method for
providing enhanced quality of service for a client device that accesses a
network via a single connection to a Point Of Presence server.
Many users connect to the Internet, also known as "world wide web", via a
switched communication line coupled to a dial-up modem. As used herein the
term Internet refers to a large distributed network comprised of
interconnected processors (servers) and routers wherein data passes
through several routers, before reaching its intended destination (e.g.,
an applications server). For Internet access devices, such as a personal
computer or so-called "web television" device, the entry point to the
Internet is usually a Point Of Presence (POP) terminal server. From the
POP terminal server, the user of the device can access any applications
server in the network.
Today, many users are attempting to transfer to their web access devices,
large amounts of data or files, such as Hyper Text Markup Language (HTML)
text, graphics or animated image files using formats such as, Graphics
Interchange Format (GIF) files, real time audio files or multimedia (video
and audio) files. These files are typically very large, bit-intensive
files. As used herein the term "download" refers to a file transfer from
the applications server through the Internet or other large distributed
network, to the POP terminal server and from the POP terminal server to
the user.
One of Internet users' most common complaints is the delay experienced by
dial-up modem Internet users in their attempt to download bit-intensive
files to their web access devices. The aforementioned delay is sometimes
so severe that the process of downloading audio and video information is
frustrating to the point that dial-up modem Internet users end up
terminating the process before most of the file is downloaded. In summary,
almost all Internet access providers are unable to guarantee to their
subscribers a given (QOS) Quality of Service (i.e., performance
requirement for delivery of data within a given time window).
In response to this problem, many solutions have been proposed to guarantee
a given QOS for Internet service. For example, some systems engineers have
suggested that timely data delivery over the Internet can be achieved by
significantly increasing the bandwidth of the links in the shared IP
backbone. Unfortunately, even an over-engineered Internet backbone may not
result in smooth transfers of real-time information (e.g., audio and
video) due to the inherent nature of the IP protocol itself which allows
only "best effort" attempts at data delivery.
Another technique for solving the delays inherent in transferring data from
an applications server to a web access device involves the use of an
additional, separate plain old telephone service (POTS) circuit-switched
connection to couple the applications server to the web access device.
While useful, this technique requires a user to have two individual
telephone lines and two separate modems, both of which being available at
once. Unfortunately, most users do not have this capability. Thus, a
problem of the prior art is the lack of an efficient, cost-effective and
practical method for significantly reducing delay over a connection that
couples a web access device to an applications server.
SUMMARY OF THE INVENTION
We have realized that for a connection that couples a web access device to
an applications server via a POP terminal server, the leg of the
connection that extends from the web access device to the POP terminal
server is more often than not congestion-free while the leg of the
connection that couples the POP terminal server to the applications server
is prone to heavy traffic density that causes intolerable delay for large
file transfer.
The present invention is directed to a method and a system for quicker
transfer of information from an applications server to a web access
device. The quicker information transfer is accomplished by bypassing the
typically congested IP backbone to establish a separate connection that
extends from the applications server to a Point Of Presence (POP) terminal
server that is coupled to the web access device. The connection may be for
example, a circuit switched ISDN connection associated with another
network that is logically separated from the backbone of the IP network.
Other alternate paths for such a connection include a logical
communications path within an Asynchronous Transfer Mode (ATM) network
with QOS feature, a logical communications path within another IP network
that is RSVP-enabled internally between routers, a link within a Frame
Relay network, or simply a communications path within any other
packet-switched or circuit-switched network that may be less congested.
According to a feature of the present invention the request and data paths
are separated within the network so that there are two distinct
unidirectional paths, one from the Point Of Presence (POP) terminal server
to the web applications server via the IP backbone using IP protocols, and
the other path being, for example, a circuit-switched ISDN connection
coupling the applications server to the POP terminal server.
According to one embodiment of the present invention, an applications
server to which a query is directed by a web access device, communicates
with a Generalized Address Resolution Protocol Server (GARPS) resident
within the IP network to determine the optimum path for downloading
requested data to a particular POP terminal server. The GARPS maps IP
addresses to a subnet address, which is a combination of a selected
subnetwork (POTS, ISDN, ATM, RSVP network, etc.) and the address within
that network for the machine identified by the IP address. According to
the present invention, any device on the network can use the GARPS find
out the best subnet to choose when setting up a QOS path to another
device.
According to another feature of the present invention, the optimum subnet
is chosen based on several key factors, such as: a) the ability to support
the QOS desired by the user on that network; b) the current traffic load
on that subnetwork; c) the price/payment that the user is willing to pay
for that service; and d) the time of day variations that the subnetworks
experience.
According to yet another feature of the present invention, the optimum
subnet will be made available by the GARPSes, and its identity will be
stored in a database that will be dynamically updated based on the above
criterion. In the event that none of the alternate subnets are available,
the default IP backbone network would be used.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 depicts the architecture of the system of the present invention.
FIG. 2 depicts a flow chart of the method according to the present
invention.
FIG. 3 depicts the interconnection of the Generalized Address Resolution
Server according to the present invention.
DETAILED DESCRIPTION
When downloading large data files from an applications server on the
Internet, most of the delay is due to the delays in the network as the
file is transferred in packets from the applications server to the Point
Of Presence (POP) terminal server. Furthermore, these delays are not very
predictable, which means that the user simply looks at his screen and
waits until more data is transferred. Once the data reaches the POP
terminal server, the delay is relatively short and very predictable based
on the file size.
Connection Management Information Transfer
One technique for overcoming this bottleneck (which has been disclosed in
an earlier application by some of the same inventors, entitled "Method of
Transferring Connection Management Information in World Wide Web Requests
and Responses," Ser. No. 08/744,232, and which is hereby incorporated by
reference as if repeated herein in its entirety, including the figures)
transfers connection management information between Web clients and
servers by means of messages incorporated in optional header fields of the
HTTP protocol header; the HTTP protocol being protocol through which web
clients and servers communicate with each other. These fields in an HTTP
message carry information related to connection management that is used
for direct communication between the client and server along alternative
paths that provide a QOS greater than what can be achieved over the
Internet.
The connection management information, incorporated within optional header
fields of the HTTP protocol header, includes addressing information which
specifies addresses on a subnetwork, such as ATM addresses, ISDN or POTS
E.164 numbers, as well as an alternate Internet Protocol (IP) address that
may be used to communicate over non-persistent or switched paths. In
addition, the addressing information includes the subnetwork type in which
the addresses are applicable. Examples of the latter may include the IEEE
802 family of networks (e.g., Ethernet or token-ring), Emulated LAN (ELAN)
ATM, ISDN, FR (Frame Relay), and X.25 virtual circuit networks. Further,
for those subnetworks for which QOS can be controlled, such as ATM or FR
networks, the connection management information transmitted in the header
fields can also include QOS information such as a required bandwidth,
maximum variance of packet delay, maximum packet loss, and preferred
Socket type (e.g., datagram (UDP), or stream (TCP)).
The optional headers in the HTTP message containing the connection
management information may be inserted by a server, a client, or an
intermediate system on behalf of the client, the latter referred to as a
proxy on the World Wide Web (WWW). The HTTP messages are initially
delivered over a router network (e.g., the Internet), using multiple
router hops rather than a direct connection between the communication
endpoints. The Web client or server endpoint then uses the additional
addressing information contained in the header to establish a direct
connection to the other endpoint to provide services (e.g., the delivery
of information) on the direct network connection. As an alternative, the
address information provided to the other end can be of an intermediate
system (IS), such as the proxy of a specialized router node, and rather
than a direct connection between the two endpoints, one of the original
endpoints may establish a direct connection on the subnetwork with the
intermediate system. When the underlying subnetwork is capable of acting
upon QOS information, QOS information included in the header may be used
by either a client, or a proxy acting on behalf of the client. A program
running on the client or proxy interprets the QOS response information,
and indicates with what QOS a direct connection (known as a cut-through)
between client and server is to be undertaken.
Generalized Address Resolution Server
One embodiment of the present invention to reduce the bottleneck in
transferring large amounts of data over the Internet (or any large
distributed network) uses a Generalized Address Resolution Protocol Server
(GARPS) resident within the network by which other devices in the network
can obtain routing information for downloading data. The GARPs maps IP
addresses to a "subnet address," which is a 2-tuple of the subnetwork
(phone, ISDN, ATM, RSVP, network, etc.) chosen and the address within that
network for the machine identified by the IP address. Any device can use
this to find out the optimum (i.e., the most appropriate) subnetwork to
choose when setting up a QOS path to another device.
According to the present invention, the optimum subnetwork is a function of
the following criterion:
a) ability to support the QOS desired by the user on that network;
b) traffic load on that subnetwork;
c) price/payment for that service (which could be negotiated with the
client or the server) that the user is willing to pay;
d) time of day variations (load-balancing at a gross level) that the
subnetworks experience. Busy hours peak at different times on different
subnetworks.
According to the present invention the optimum subnetwork will be made
available by the GARPSes, and will be stored in a database that will be
dynamically updated based on the above criterion. In the event that none
of the alternate subnets are available, the default IP network would be
used (with a suitable warning). In the other embodiment, this information
is wrapped inside the request sent out by the client. However, the use of
the GARPs allows greater flexibility, and does not require either:
the client to know the subnet address;
the network to wrap the subnet address inside the client's request (thereby
modifying the request) using a proxy of some sort at the client POP.
Much like Address Resolution Protocol (ARP), which is used currently in the
IP world to populate mappings between IP addresses and their MAC addresses
(a kind of subnet address), GARP or proxy GARP is a protocol that dial-in
customers can use to access to any of these subnetworks directly. For
instance, if a client has another phone line connected on another PPP
interface, a GARP message could be sent out that essentially says:
map: myIP address.fwdarw.(phone-subnetwork, myPhoneNumber) the GARP message
would also specify the QOS that the subnet can support (14.4 or 28.8 kbps
etc.). Furthermore, the users can access the GARP server and its database
to create their profile, which is then used to determine the optimum
subnetwork. For example, the user can indicate that he has access to a
particular ATM network and the associated address on that network. Then,
when attempting to determine which subnetwork to download the data over,
the GARP server will examine the particular user's profile, apply the
above mentioned criteria, and depending on the right circumstances, may
choose the particular ATM network.
While the exact location of the GARP server on the network is not
important, what is important is that the destination server, which is to
download the data, must have connectivity to the GARP server. Thus, once a
request to download data is received by a destination server that has been
configured to search for an optimum alternate path, the destination server
will contact the GARP server, determine the optimum path for the
designated user or POP terminal server, and send the data over that route.
The present invention therefore reduces the delay inherent in the transfer
of the file from the applications server to the POP terminal server by
establishing an alternate connection, such as a circuit switched
connection, from the web applications server to the POP terminal server
outside the IP network and over an alternate infrastructure instead of via
the IP backbone.
The present invention enables service providers the ability to guarantee
performance for all user requested information--HTML, text, graphics,
images and real-time multimedia content. The present invention utilizes
the existing POTS infrastructure to deliver content taking advantage of
the fact that peaks of voice traffic and Internet traffic occur at
different times, i.e., voice traffic peaks occur during business hours
(9-5), whereas Internet traffic peaks occur during the early evening hours
(7-10 p.m.). As a result, there exists excess capacity in the circuit
switched network when the IP backbone is full, which is precisely when
attempts to download large amounts of data become extremely slow.
In comparison to the present invention, most Internet applications (such as
Real Audio, for example) the signaling (e.g., the number to dial) is done
expressively by the application. In accordance with the present invention,
the signaling information is handled without requiring the knowledge of
the user or application.
System Overview
Referring to FIG. 1, the user 1 accesses the Internet via a dial-up
connection 2 to a Point Of Presence terminal 3 server within the network
4. Through a series of routers 5, 6, the user can access an applications
server 7 with data of interest to the user 1.
Once the user 1 has decided to download the data, according to the present
invention, the applications server 7 that is to download the data
determines the appropriate alternate path by which to transfer the
information rather than the through the IP backbone. The applications
server determines this by contacting the GARP server's database, which
lists the optimum subnetwork for the specified user/POP terminal server.
The alternate path could be a circuit switched path via ISDN lines, a QOS
ATM network, another IP network that is RSVP enabled internally between
routers, Frame Relay or other networks, for example.
Alternatively, the applications server can be instructed to send the data
via a specified alternate path. This instruction can either be embedded in
the request for the data (i.e., as in the connection management
information transfer technique described above) or as a configuration
operation (by the GARS described above). The applications server 7 then
establishes the connection, e.g., an ISDN connection 9 to the circuit
switched network 8. The circuit switched network then establishes a
connection to the POP terminal server 3 via another ISDN line 10. At the
POP terminal server 3, the data is multiplexed with IP data being sent to
the user.
There are three approaches to connecting the alternate network 8 to the POP
terminal server 3. In the first approach, the ISDN connection 10 to the
POP terminal server 3 can be made as one of the many dial-up connections
to the POP terminal server 3. The POP terminal server 3 can process many
different telephone calls at once at its front end. Assuming there are
lines available, the ISDN connection 10 from the circuit switched network
8 can be made via one of these telephone lines. The POP terminal server 3
then simply routes the incoming ISDN connection 10 to the user connection
11. Alternatively, this call could be passed to the first router 5, which
would then reroute the call back to the POP terminal server 3 and on to
the client 1.
One must use caution, however, in assigning too many telephone lines for
this purpose as this connection is the only "entrance" to an IP service
provider's "store." As many Internet service providers have learned, users
become disillusioned with an IP service provider if they cannot even
access it. Therefore, in attempting to create a solution to long delays in
downloading data, one cannot create problems for users attempting to logon
by allowing users already logged on to occupy two connections to the POP
terminal server. This essentially would cut the number of access lines in
half, assuming all users were downloading data. Thus, where limited
telephone lines are available, this solution may be less than optimal.
A second approach is to connect the ISDN line 10 to the back end of the POP
terminal server 3. The POP terminal server 3 then would connect the user
connection 3 to the ISDN connection 10. While slightly more complex than
the previous approach, this approach does not potentially use up scarce
resources of the POP terminal server 3. This connection would be similar
to any of the router 5 connections to the POP terminal server 3.
Yet another alternative if no incoming lines to the POP terminal server 3
are available is to connect the alternate network to another server within
the IP network that is directly connected (or logically "closely
connected") to the POP terminal server 3 so that there is very little
delay added by the connection from the other server to the POP terminal
server 3. This can also be referred to as the nearest neighbor. Usually,
most delays are due to routers and concentrators that are overloaded with
traffic. If you can avoid the traffic jams, you can significantly reduce
the delay, especially where there is only one transfer as in this case.
This approach avoids these "traffic jams" at these overloaded junctions.
If the circuit switched network is itself overloaded, then the data path
to the POP terminal server occurs in the normal fashion.
Depending on the type of network, one can expand this concept to include
those closest servers that are only a few hops away. By setting a maximum
limit on the number of routers between the POP terminal server 3 and the
server that the circuit switched network is overloaded, one can at least
limit the potential delays involved.
By improving the download time on the network side, the present invention
enables an IP service provider to make available faster connections to the
POP terminal server 3 as the speed increase will be actually noticed by
users, rather than being an insignificant improvement in performance. If
users can actually realize an improvement in performance, they are more
likely to be willing to pay for it. As a result, tier pricing for higher
capacity access lines can be significantly increased.
Turning to FIG. 2, the method 20 of the present invention is as follows.
When the user requests certain data to be transferred to him or her, the
request (i.e., the uplink information) comes through as usual using IP
over a dialup connection to the terminal server at the access POP, as
shown in step 21. The POP then routes this request over the network
infrastructure to the Home-Page server or other URLs as applicable (see
step 22). The Home-Page server(s) establishes an alternate QOS path, such
as an ISDN path, back to the POP on a need-basis or pre-established basis,
as in step 23. The Home-Page server uses an established QOS protocol to
send the information back to the POP over the QOS-enabled network. (see
step 24). The POP routes this information back to the user on the existing
dialup connection, as shown in step 25.
Turning to FIG. 3, a Generalized Address Resolution Protocol Server 31 of
the present invention is coupled to a Local Area Network (LAN) 32, to
which a POP terminal server 33 could also be coupled. The POP terminal
server 33 is coupled to the user 34, and is the route by which the user 34
accesses the Internet 35. The POP terminal server 33 has n connections,
one of which is assigned to the user. The user accesses the POP terminal
server using PPP protocol with a dial-up modem. The POP terminal server
also has an alternate connection to it, such as an ISDN line, an QOS
network, etc. When data is sent to the user 34 over the alternate
connection, the POP terminal server uses the loop-back output (LO) to
return the data to the POP terminal server, as if the data was coming from
the LAN 32 or Internet 35. The data is then transmitted to the user in the
normal fashion.
When the destination server desires to download data to the user, it
contacts the GARP server 31, which indicates the best alternate path to
the user. The GARP may be anywhere on the network, not just as shown as in
FIG. 3. The GARP server 31 will maintain a database 37 of the appropriate
return path in view of the above criterion, which is constantly in a state
of flux. The return path indicator will instruct the destination server 38
to establish a path from itself 38 to the POP terminal server 33 along the
designated path. As a result, the data downlink will be established
independently of the IP network, if a QOS path exists.
Top